| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
|
|
|
|||||||
|
|||||||||
Rapid Responses to:
|
|
Rapid Responses published:
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif){kind=icon}
Re: Errors in confidence intervals
Errors in confidence intervals
|
|
|||
|
Ian A. Young, Physical Therapist, Spine and Sport, Savannah, Georgia, Lori A. Michener, Joshua A. Cleland, Arnold J. Aguilera, Alison R. Snyder
Send rapid response to journal:
youngian{at}spinesport.org Ian A. Young, et al.
|
We thank Dr Thorpe for bringing up a few very important issues with Table 3 in this article.1 She was correct in her assumptions about the nonsymmetrical 95% confidence intervals (CIs). Please refer to the revised Table 3 where the corrections address transposition errors, inconsistent use of 100th versus 10th decimal points that are now consistent, and incorrect calculation of 95% CIs. All values and 95% CIs have been verified for accuracy, and the correct values appear in the revised Table 3. The 95% CIs for adjusted effect sizes at 4 weeks were reported incorrectly in the text. The corrected Results section appears in the data supplement. In addition, because of the changes to the 95% CIs, the second paragraph in the Discussion section—on precision of point estimates—has changed. The corrected paragraph appears in the data supplement. Ian A. Young, Lori A. Michener, Joshua A. Cleland, Arnold J. Aguilera, Alison R. Snyder Ian A. Young, PT, MS, OCS, SCS, CERT MDT, is Physical Therapist, Spine and Sport, Savannah, Georgia, and Affiliate-Associate Professor, Department of Physical Therapy, Virginia Commonwealth University–Medical College of Virginia Campus (youngian{at}spinesport.org) Lori A. Michener, PT, PhD, ATC, SCS, is Associate Professor, Department of Physical Therapy, Virginia Commonwealth University–Medical College of Virginia Campus Joshua A. Cleland, PT, PhD, OCS, FAAOMPT, is Associate Professor, Department of Physical Therapy, Franklin Pierce University, Concord, New Hampshire; Physical Therapist, Rehabilitation Services, Concord Hospital, Concord, New Hampshire; and Faculty, Regis University Manual Therapy Fellowship Program, Denver, Colorado Arnold J. Aguilera, MD, is Neurologist, Neurology Associates, Fredericksburg, Virginia Alison R. Snyder, PhD, ATC, is Assistant Professor, Athletic Training Program, A.T. Still Univeristy, Mesa, Arizona Reference 1 Young IA, Michener LA, Cleland JA, et al. Manual therapy, exercise, and traction for patients with cervical radiculopathy: a randomized clinical trial. Phys Ther. 2009;89:632–640. |
|||
|
|
|||
|
Donna L Thorpe, Professor in Physical Therapy Loma Linda University
Send rapid response to journal:
dthorpe{at}llu.edu Donna L Thorpe
|
I appreciate the discussion in the article by Young et al that focused on the meaning and precision of confidence intervals regarding effect sizes seen in the study. I think this focus on the meaning of the data is extremely useful. However, it would have been more helpful if the confidence intervals had “fit” the point estimates for the Numeric Pain Rating Scale (NPRS) and the Neck Disability Index (NDI). Unfortunately, the point estimates given for the NPRS and the NDI are not at the midpoint of the confidence intervals—a necessary requirement of confidence intervals of mean differences. After looking at them carefully, it is fairly easy to guess that the confidence intervals were probably constructed around a negative value for the point estimates (negative differences between group means), rather than the positive value given. It appears that this mistake also may have been repeated 8 of the 32 confidence intervals in Table 3. A second type of error (or conflict) in the data is seen in the 4-week unadjusted mean difference between groups for the Patient-Specific Functional Scale and the Global Rating of Change Scale. The intervals stated do not include the value “0” and, therefore, indicate the differences are significant. However, the associated P value clearly indicates the differences are not statistically significant. I suspect the lower limit value for the confidence interval may have been negative rather than positive. Donna Thorpe, PT, DrPH |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
Rapid Responses: